The 3 best predictors of staff performance

In recruitment, the easiest thing to do is the same thing you’ve always done.

But whether you’ve had a staff member resign, or you’re hiring for new positions, it’s always worth double-checking your selection process to consider what you could be doing better.

For example, to have the greatest chance of picking the right person for the job, it’s worth checking that you’re using the most effective tools to assess your candidates in the first place.

Don’t leave it up to chance

You probably wouldn’t make a big decision about one of your organisation’s core work areas without doing a bit of research first. Choosing selection tools should be no different.

A widely-referenced research paper by US Professors Frank Schmidt and John Hunter provides some excellent, and at times surprising, insights into the data behind different selection tools and recruitment processes. The pair analysed an enormous 85 years of research to identify the most effective methods for hiring people who will excel in their roles – and the least effective.

Here are some of their key findings:

Best avoided: unstructured interviews

Tell me about yourself. What’s your greatest strength? What’s your greatest weakness?

These are long-established, popular, but unstructured (or “open-ended”) interview questions. They may seem tempting if you’re short on time or you like to go with your gut instincts.

But here’s why you should think twice: unstructured interviews are significantly less effective than structured interviews, and from Schmidt and Hunter’s research were only the 9th most effective method of predicting a candidate’s future success.

They found that unstructured interviews have a predictive validity of just 38% – that is, a candidate’s performance in an unstructured interview only matches up to how well they go on to perform in their role in 38% of cases.

More organisations and HR managers are becoming familiar with the pitfalls of unstructured interviews when planning their hiring processes and interviews – like Google’s HR boss Laszlo Bock.

Top predictors of success

According to Schmidt and Hunter’s research, the top three predictors of success for a future candidate are:

1. Work sample tests, or giving candidates a similar piece of work to do, then comparing how each performs in the task. Although this method comes in at number 1, it’s still only accurate in predicting job performance 54% of the time when used alone. This method can’t, for example, evaluate interpersonal skills, or the ability to deal with uncertainty.

2. Structured interviews – that is, interviews that ask each candidate the same set of standardised and job-specific questions, with clear criteria to assess the quality of responses. These are significantly more successful than unstructured interviews in predicting a how well a candidate will fare in their future job.

3. General Mental Ability (GMA) tests, which basically aim to evaluate how “smart” someone is. The problem with these kinds of tests is that they can discriminate against non-white, non-male test takers. Despite this, research shows they are the most accurate in predicting job performance, as well as the added benefit that they can be used for all jobs, and are very cost effective.

The winning combination: tests and structured interviews

Probably the most useful takeaway from the research is that a combination of methods is by far and away your best bet to maximise your chances of picking a great candidate. And Schmidt and Hunter recommend always including a GMA test.

Their research shows that, when combined with a GMA test, the best methods to use when you’re hiring are:

  • Work sample tests
  • Structured interviews
  • Integrity tests – which test how honest a candidate is in potential work situations
  • Conscientiousness tests – which test a candidate’s self-discipline, ability to organise themselves, plan, set goals and follow rules.

For example, combining a general cognitive ability test with an assessment of conscientiousness will paint a more accurate picture of how a candidate will perform in their job than either single method will alone.

How all 19 methods fare

Here’s how the complete list of methods that Schmidt and Hunter assessed fared, in order of most to least effective in predicting future job success:

  1. Work sample tests: 54%
  2. Structured interviews: 51%
  3. GMA tests: 51%
  4. Peer ratings: 49%
  5. Job knowledge tests: 48%
  6. Training and experience behavioral consistency method: 45%
  7. Job tryout procedure: 44%
  8. Integrity tests: 41%
  9. Unstructured interviews: 38%
  10. Assessment centres: 37%
  11. Biographical data measures: 35%
  12. Conscientiousness tests: 31%
  13. Reference checks: 26%
  14. Job experience in years: 18%
  15. Training and experience point method: 11%
  16. Years of education: 10%
  17. Interests: 10%
  18. Graphology (handwriting analysis): 2%
  19. Age: –1%

The list shows the percentage of the time a particular method accurately predicts future job performance when used on its own – once combined, the numbers increase.

Finding the right approach for your organisation

The method or combination of methods you end up choosing for your recruitment process will of course depend on your organisation’s resources.

But even the most cash-strapped NFP can start developing structured interviews, or think about including a work sample test as part of a recruitment process.

Whatever you do – don’t forget the importance of evaluating how things go! Experiment where possible, test new ways of hiring people, and make sure to build up your organisation’s own data on what works.

How does your organisation assess candidates? Do you experiment with different options? We’d love to read about your experiences, good or bad, in the comments below!

This post is based on Hunter and Schmidt’s original research paper. You can also read about the research in this Wired article.

One Coment, RSS

  • August 7, 2017 @ 8:22 pm Reply

    Thanks for your article. Schmidt and Hunter’s research on job performance provides correlations for different selection methods not percentages.
    1.0 is a perfect positive correlation (very strong relationship) with performance and 0.0 is no correlation.
    0.54 is a moderate positive correlation meaning it has moderate predictive ability.

    Your email address will not be published. Required fields are marked *

  • Name *
  • Website